Circular RNA circFGD4 suppresses gastric cancer progression via modulating miR-532-3p/APC/β-catenin signalling pathway

Gastric cancer (GC) ranks as the fifth most common malignant gastrointestinal tumours, and it is the major cause of cancer mortality not only in China but also in the world [1,2]. At present, most patients with GC are still diagnosed at an advanced stage, and the 5-year overall survival rate of GC patients remains lower than 30% [3,4]. Despite the diagnosis and therapeutic strategies continue to be enhanced, rapid proliferative capacity, lymphatic/hematogenous metastasis, and direct invasive behaviour are key factors leading to poor prognosis in GC patients. Since the occurrence and development of GC is a complex biological process, and thus far, the current molecular mechanism of GC progression has not been clearly understood [5]. The critical situation of GC diagnosis and therapy has prompted us to develop effective therapeutic strategies for anti-GC therapy and to search for valuable biomarkers to monitor GC progression.

Circular RNAs (circRNAs) are a category of non-coding RNAs (ncRNAs) formed by a closed continuous loop structure [6,7]. circRNAs are highly stable and are cell type- and tissue stage-specific, indicating that most of the circRNAs are independent of the host gene. Most recent studies have confirmed that circRNAs have multiple molecular functions, such as miRNA sponges, protein translation, and circRNA–protein interaction [8,9]. Moreover, circRNAs also play critical roles in several physiological and pathological processes, including cell cycle progression, autophagy, proliferation, invasion, metastasis, and carcinogenesis [10,11]. Nevertheless, the molecular characterizations and mechanisms of most circRNAs in human GC have not been elucidated and need further investigation. Here, circRNA microarray analysis was performed in five pairs of GC tissues to evaluate the expression of circRNAs in GC tissues and matched normal samples. Then, we screened and identified the expression and function of hsa_circ_0000390 derived from FYVE, RhoGEF and PH domain-containing 4 (FGD4) in GC and examined the underlying mechanism of hsa_circ_0000390 in GC carcinogenesis and development.

In our present study, we designed the functional and molecular experiments to examine the functional role of circFGD4 in GC. circFGD4 was markedly down-regulated in human GC tissues and cell lines and suppressed the progression and metastasis of GC cells through sponging miR-532-3p and inactivating the APC–β-catenin signalling. Importantly, low expression levels of circFGD4 were associated with poor tumour differentiation, lymph node metastasis, and poor prognosis of GC patients. The data suggest that circFGD4 serves as a tumour suppressor molecule involved in GC progression and provides an independent prognostic biomarker and target for the management of GC.

All human GC tissues and noncancerous tissues from January 2013 and August 2014 were collected from the First Affiliated Hospital of Chongqing Medical University and stored in liquid nitrogen. The final diagnosis was histopathologically confirmed by two pathologists. The clinical–pathological features of all patients were recorded and are summarized in Table 1. The present study was approved by the Ethics Committee of the First Affiliated Hospital of Chongqing Medical University (Chongqing, China). Written informed consent was provided by all the participants, which agreed to participate in the present study.

Total RNAs isolated and extracted from the five human GC tissues and paired non-cancerous tissues were used for microarray analysis. RNA amplification, labelling, and hybridization with microarrays were the same as previous studies [12]. Briefly, CapitalBio circRNA Amplification and Labeling Kit (CapitalBio Co., Ltd, Beijing, China) were applied to amplify and transcribe the RNase R enriched circRNA into fluorescent cRNA with 100–500 ng purified circRNAs. The labelled RNAs were hybridized onto the Human CircRNA Array v2.0. Array hybridization was performed in an Agilent Hybridization Oven overnight at a rotation speed of 20 rpm and 42°C (0.2% SDS, 2× SSC at 42°C for 5 min, followed by 0.2× SSC at 37°C for 5 min). The signals were scanned using the Agilent G2565CA Microarray Scanner, and array images were analysed with the Agilent Feature Extraction software (v10.7). The raw data quantile normalization and data analysis were analysed with Agilent GeneSpring software.

Total RNA was isolated and extracted from human GC tissues and cells using TRIzol Reagent (Takara, #9108, China). The treated RNA was reverse transcribed into the first-strand cDNA by a PrimeScript RT Reagent Kit (Takara, #RR037A) and random primers following the Kit’s protocols. The circFGD4 primers designed across the circularization site were designed by Geneseed (Guangzhou, China). All primers (GAPDH, circFGD4, FGD4, APC, Ki-67, PCNA, E-cadherin, and Vimentin; or miR-532-3p and U6) were designed and synthesized by Sangon Biotech (Shanghai, China) or Ribobio (Guangzhou, China), and the primer sequences used are shown in Table 2. The expression of circFGD4, miR-532-3p, and related mRNA in GC tissues and cells was quantified by qRT-PCR using TB Green Premix Ex Taq (Takara, #RR820A). GADPH and U6 were employed as mRNA and miRNA endogenous controls, and the 2^−ΔΔCt formula was conducted to analyse the relative expression levels. The PCR products were investigated using 2% agarose gel electrophoresis. The bands were separated by electrophoresis at 120 V for 20 min and were examined by UV irradiation. Then, the PCR products were further identified through Sanger sequencing by Sangon Biotech.

Human GC cell lines (MKN28, AGS, SGC7901, MKN45, MGC803, and BGC823) were purchased from the Type Culture Collection of the Chinese Academy of Sciences (Beijing, China). The human embryonic kidney cell line (HEK-293T) and the normal gastric mucosal epithelial cell line (GES-1) were donated by the Molecular Tumor and Epigenetics Laboratory from the First Affiliated Hospital of Chongqing Medical University. All cells were cultured in DMEM or RPMI 1640 medium (Gibco, Carlsbad, U.S.A.) supplemented with 10% foetal bovine serum (FBS, Gibco, Grand Island, U.S.A.) and stored in a humidified atmosphere of 5% CO₂ at 37°C. For cell treatments, 2 μg of total RNA was incubated for 30 min at 37°C using 3 U/μg of Ribonuclease R (RNase R; Epicentre Technologies, U.S.A.). Next, GC cells were treated with 2 μg/ml Actinomycin D (Sigma-Aldrich, USA) for 0, 6, 12, 18, and 24 h, respectively. Relative RNA levels of circFGD4 and parental gene FGD4 were performed by qRT-PCR and normalized to the values measured in the 0 h group.

To construct circFGD4-overexpressing vector, a 345 bp cDNA of circFGD4 was cloned into the pHBLV-CMV-crRNA-EF1-T2A lentiviral or plasmid vector purchased from Hanbio Biotech (Shanghai, China). To knockdown circFGD4, three individual circFGD4 siRNAs (siRNA1, siRNA2, and siRNA3) targeting the back-splice junction site of circFGD4 and a siRNA negative control (si-NC) were designed. Based on qRT-PCR results, siRNA1 was significantly the most effective siRNA and was used to construct the siRNA plasmid, named shRNA1, and si-NC was used to construct the sh-NC plasmid (GenePharma). The miRNA negative control (miR-NC), miR-532-3p mimics, and miR-532-3p inhibitors were designed and synthesized by Ribobio (Guangzhou, China). The nucleotide sequences are listed in Table 2. All oligonucleotides and overexpressing- or knockdown-plasmids were transfected into GC cell lines using Lipofectamine 2000 Reagent (Invitrogen, Carlsbad, U.S.A.). After transfection for 48–72 h, cells were harvested for further experiments.

GC cell viability detection was employed using the CCK-8 Kit (Dojindo, Kumamoto, Japan) following the manuals from the manufacturer. Briefly, about 10³ transfected GC cells were seeded into 96-well plates, and 10 μl CCK-8 was added to each well for 3 h at 37°C with 5% CO₂. The optical density value was evaluated in 0, 24, 48, 72, and 96 h at 450 nm using a Microplate Reader. Next, approximately 200 transfected GC cells were seeded into 24-well plates and cultured for 14 days regarding colony formation assay. The colonies were fixed with 4% paraformaldehyde and stained using 0.1% Crystal Violet, and visible colonies were photographed and manually counted.

Proliferating cells were assessed using an EdU Labeling/Detection Kit (RiboBio, China) following the Kit’s instructions. The treated GC cells were seeded in 96-well plates at 1 × 10⁴ cells per well and cultured to the logarithmic growth phase. The abovementioned GC cells were treated with EdU solution and Apollo staining reaction solution. Cell nuclei were stained with Hoechst 33342 reaction solution. The percentage of EdU positive cells was evaluated using a confocal laser scanning microscopy (LSM800, Carl Zeiss AG, Germany), and images were obtained and counted.

For cell transwell assay, the serum-free medium containing 4 × 10⁴ transfected cells was placed into the 24-well transwell chambers (8 μm pore size, Corning Costar, U.S.A.), and complete medium containing 10% FBS was added to the lower chambers. After incubation for 24 h, the migratory cells at the lower surfaces of chambers were fixed using methanol and stained using 0.1% Crystal Violet solution, and next photographed. The transfected target cells were evenly plated into a six-well plate, and the cells were cultured to 90% confluency after 24 h for wound healing assays. A 10 μl sterile tip perpendicular to a six-well plate was used for scratches and photographed at 0 h. Then, the cells were cultured in the serum-free medium, and images were obtained after 24 h.

RNA FISH probes for the circFGD4 and miR-532-3p sequences were designed and synthesized by GenePharma. GC cells were seeded and fixed in 24-well plates containing climbing pieces. Cy3 probes were specific to circFGD4, and FAM probes were specific to miR-532-3p. 4′6-Diamidino-2-phenylindole (DAPI; Life Technologies, Carlsbad, U.S.A.) was applied to label cell nuclei. The probe signals were measured by using Fluorescent In Situ Hybridization Kit (GenePharma) following the Kit’ s guidelines. The fluorescent signals were determined by using a confocal microscope (LSM800, Carl Zeiss AG, Germany). The sequences of circFDG4 and miR-532-3p probes are listed in Table 2.

Briefly, GC cells were plated onto 24-well plates with climbing pieces and transfected with the circFGD4 overexpression plasmid, the shRNA1 plasmid, and some of the cells were also cotransfected with miR-NC or miR-532-3p mimics + circFGD4 overexpression plasmid and miR-NC or miR-532-3p inhibitors + circFGD4 shRNA1 plasmid separately for 48 h. The cells were incubated with APC, E-cadherin, and Vimentin (Abcam, Burlingame, U.S.A.) primary antibodies at 4°C overnight, followed by incubating the corresponding fluorescent secondary antibodies. DAPI was used to stain the cellular nuclei, and all transfected cells were detected and observed using a confocal microscope (LSM800, Carl Zeiss AG, Germany).

RIP-binding protein immunoprecipitation assay was carried out in AGS cells by a Magna RIP RNA-binding protein Immunoprecipitation Kit (Millipore, Billerica, U.S.A.) and human anti-Argonaute2 antibody (AGO2, Cell Signaling Technology, U.S.A.) following the manufacturer’s protocols. Briefly, 1 × 10⁷ AGS cells stably transfected with miR-NC or miR-532-3p mimics were washed by ice-cold PBS and lysed in an equal volume of RIP lysis buffer. The cell lysates were incubated with negative control anti-AGO2 antibody or rabbit IgG-coated magnetic beads and rotated at 4°C overnight. After treating the cell lysates with proteinase K buffer, the RNA fraction was purified and extracted to detect the AOG2 protein, circFGD4, and miR-532-3p.

For circFGD4 pulled down miRNAs, AGS cells (> 1 × 10⁷) stably expressed circFGD4 were harvested, lysed with lysis buffer, and sonicated. To generate probe coated beads, the biotinylated circFGD4 probe (control: oligo probe) was incubated with streptavidin magnetic beads (Invitrogen, U.S.A.) at room temperature for 2 h. The cell lysates were incubated using the circFGD4 probe at 4°C overnight. Next, the treated beads were washed, and the RNA complexes were eluted, purified and extracted with TRIzol (Takara) for qRT-PCR. Biotinylated probe sequence of circFGD4 was designed (GenePharma, China) and is showed in Table 2.

Luciferase reporter plasmids (pGL3- wild-type (WT) or -mutant-type (MUT) circFGD4 fragment and pGL3- WT or -MUT APC fragment) were designed and synthesized by GenePharma (China). HEK-293T cells grown in a 96-well plate were cotransfected with corresponding plasmids containing pGL3- WT or -MUT circFGD4 and APC fragment and miR-532-3p mimics or control miR-NC using Lipofectamine 2000 (Invitrogen, U.S.A.). After culturing for 48 h, the luciferase values based on firefly and Renilla were detected by the Dual-Luciferase® Reporter Assay System (Promega, Madison, U.S.A.). Relative luciferase activity was calculated as the ratio between the firefly and Renilla luciferase values for each sample.

HEK-293T and GC cells were cultured for 12–24 h and then cotransfected with the miR-NC, miR-532-3p mimics, miR-532-3p inhibitors, miR-532-3p mimics + circFGD4 overexpression plasmid, and miR-532-3p inhibitors + circFGD4 shRNA1 plasmid respectively combined with TOPFlash, FOPFlash and Renilla luciferase plasmid (Addgene, Cambridge, U.S.A.). After 48 h of transfection, cells were lysed by using lysis buffer, and then stimulated with 20 mM LiCl- or 15% Wnt3A- conditioned medium for 6 h. The luciferase activities were detected with Luciferase Reporter Gene Assay Kit (Promega) [13]. All luciferase activities were normalized to Renilla luciferase activity and are presented as relative luciferase activities.

Proteins extraction from GC cells were conducted using RIPA buffer containing complete protease and phosphatase inhibitors for 30 min on ice and sonicated. Also, Nuclear and Cytoplasmic Protein Extraction Kits (Beyotime, China) were utilized to isolate and extract the cytoplasmic and nuclear proteins separately. The equal quantities of protein extractions were separated on 10% SDS-PAGE and then transferred onto polyvinylidene fluoride (PVDF) membranes (Bio-Rad, CA, U.S.A.). These membranes were blocked with quick blocking solution and incubated with the specific primary antibodies against PCNA, Ki-67, Vimentin, E-cadherin, N-cadherin, APC, β-catenin, β-actin, GAPDH, and Lamin B1 (Abcam) at 4°C overnight. The membranes were washed with TBST buffer and incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies at room temperature for 2 h. After washing, samples were visualized using the enhanced chemiluminescence (RCL) reagent (Bio-Rad, U.S.A.).

Four-week-old BALB/c female nude mice were purchased from the National Laboratory Animal Centre (Shanghai, China). Approximately 5 × 10⁶ MGC803 cells transfected with lentivirus overexpressing circFGD4 and negative control vector (NC) were administrated into the axilla of each mouse by subcutaneous injections (five mice per group). Nude mice were killed by intraperitoneal injection of a 3-fold anesthetic dose of sodium pentobarbital (150–200 mg/kg) after 4 weeks, followed by the xenograft tumours were harvested and calculated as 0.5 × length × (width)², and then used for qRT-PCR and IHC assays. For the metastasis model, 1 × 10⁷ transfected MGC803 cells in 50 μl of PBS were injected into BALB/c nude mouse (4 weeks old) via their tail veins, including the blank group, the NC group, and the circFGD4 overexpression group. After 4 weeks of feeding, the mouse was killed and examined the number and size of macroscopically visible lung metastatic nodules in nude mice. The images of metastatic nodules in mouse lungs were captured, and the metastatic nodules were used for qRT-PCR. All animal protocols were under the approval of the Animal Ethics Committee of Chongqing Medical University, and conducted at the Animal Experimental Centre of Chongqing Medical University, and carried out following the National Institute of Health Guide for the Care and Use of Laboratory Animals.

Briefly, mice xenograft tumours embedded in paraffin, sectioned into 4 μm thick, placed on glass slides and stained with IHC. After the sections were deparaffinized, rehydrated, and antigen retrieval, the sections were incubated at 4°C overnight with primary antibodies against APC, Ki-67, PCNA, E-cadherin, and Vimentin (Abcam; 1:50 dilution). Next day, after incubation with HRP-conjugated goat anti-rabbit IgG at 25°C for 1 h, then the sections were stained with diaminoaniline (DAB) and counterstained with haematoxylin. Lastly, the sections were covered with coverslips for a fluorescence microscope (Axio Imager A2, Carl Zeiss AG, Germany).

The graph plots and statistics were carried out using GraphPad Prism 8.2.1 software. P-value < 0.05 was taken as statistically significant. Statistical significance was evaluated by Student’s t-test or analysis of variance followed by multiple comparison tests. The association of circFGD4 levels with the GC patients’ clinic-pathological variables was analysed by the chi-square(χ²) test or Fisher’s exact test. Overall survival (OS) curves were performed with the Kaplan–Meier method and analysed by the log-rank test. The correlation between circFGD4 and miR-532-3p expression or circFGD4 and FGD4 expression was analysed by Pearson’s correlation coefficient analysis. All results were conducted at least three independent experiments.

To investigate the circRNA expression profile in GC tissues, we characterized circRNA transcripts in GC tissues by analysing databases from the GEO database (GSE100170 and GSE131414) and our circRNA microarray (Figure 1A). In our circRNA microarray, the clustered heat map showed that a total of 3988 dysregulated circRNAs with 2-fold change (P<0.05) were screened and identified between GC tissues and paired noncancerous tissues (Figure 1B). By excluding low-abundance circRNAs (scores are too small) from the circBase website [14], two circRNAs (hsa_circ_0000390 and hsa_circ_0007315) significantly differentially expressed in circRNA microarray were verified by qRT-PCR analysis for their expression levels in GC tissues. We found that the relatively higher hsa_circ_0000390 expression levels than the expression levels of hsa_circ_0007315 in GC tissues (Figure 1C). The qRT-PCR analysis also confirmed that the expression of hsa_circ_0000390 was lower expressed in six GC cell lines (MKN28, SGC7901, AGS, MKN45, MGC803, and BGC823) than in GES-1 cells (Figure 1D). Therefore, hsa_circ_0000390 was selected as the studied molecule.

Under RNase R treatment, the levels of FGD4 linear forms were significantly reduced, but RNase R could not digest hsa_circ_0000390 in SGC7901 cells (Figure 1E). Actinomycin D is a transcriptional inhibitor that has been used to treat SGC7901 cells. The results validated that the half-life of hsa_circ_0000390 was more than 24 h, while the half-life of linear FGD4 mRNA was lower than 6 h by qRT-PCR analysis, indicating hsa_circ_0000390 is highly stable in SGC7901 cells (Figure 1F). Bioinformatics analysis showed that hsa_circ_0000390 was located on chromosome chr12:32760890-32764217, was 345 bp long, and was formed from exons 8, 9 and 10 of FGD4; thus, we termed it circFGD4. To more rigorously validate the result of back-splicing, a convergent primer for FGD4 mRNA and a special divergent primer for the amplification of circFGD4 were designed. Extraction of cDNA and gDNA from BGC823 and SGC7901 cells for gel electrophoresis detection, and results displayed that circFGD4 could be detected only in cDNA, but no products in gDNA (Figure 1G). Also, the back-splicing junction of circFGD4 was further confirmed in PCR products by Sanger sequencing (Figure 1H). The expression level of circFGD4 was reduced in GC patients with poor tumour differentiation (P=0.033) and lymphatic metastasis (P=0.04) (Table 1). More important, low circFGD4 expression had a significantly poor prognosis of GC patients compared to patients with high circFGD4 expression (P=0.0391; Figure 1I).

To investigate whether circFGD4 levels correlate with linear FGD4 mRNA levels in GC tissues, the data from the TCGA database indicated that linear FGD4 mRNA levels had no significant difference between 408 GC tissues and 211 paired noncancerous tissues (Figure 1J). Also, there was no statistical change in the expression level of linear FGD4 mRNA at different GC stages (Figure 1K). The correlation analysis displayed that circFGD4 levels were not specifically associated with FGD4 mRNA levels in 30 GC tissues (Figure 1L). Additionally, RNA FISH assays displayed that circFGD4 was predominately localized in the cytoplasm of GC cells (Figure 1M). Together, these results suggested that lower circFGD4 expression was generally observed in GC tissues and cells and correlated with a worse prognosis in GC patients.

Given the down-regulation of circFGD4 in GC tissues and cell lines, the loss-of-function assays were applied to verify whether circFGD4 could affect the viability, colony formation, and migration of GC cells. We analysed the shRNA1 sequence of circFGD4 at back-splice junctions and designed the circFGD4-shRNA1 plasmid accordingly (Figure 2A). After shRNA1 was transfected into AGS and BGC823 cell lines for 48 h, the silencing efficiency of circFGD4 was detected by qRT-PCR (Figure 2B). CCK-8 and colony formation assays indicated that knockdown of circFGD4 expression promoted the cell viability and colony-forming abilities of AGS and BGC823 cells (Figure 2C,D). EdU incorporation assays indicated that DNA synthesis in AGS and BGC823 cells was accelerated after knockdown of circFGD4 compared with the sh-NC vector (Figure 2E). Transwell assays demonstrated that the cell migration of both AGS and BGC823 cells were prominently promoted by the circFGD4 shRNA1 transfection, which was consistent with the wound healing assays (Figure 2F,G). Additionally, IF assays indicated that the knocking down of circFGD4 increased vimentin expression and decreased E-cadherin expression in AGS cells (Figure 2H). These findings suggested that the silencing of circFGD4 expression promoted GC cell proliferation, migration, and induced EMT in vitro.

Based on circFGD4 knockdown experiments, we investigated the potential functional role of circFGD4 through gain-of-function experiments in GC cells. qRT-PCR assays indicated that overexpression efficiency of circFGD4 after the transfection was efficient in MGC803 and SGC7901 (Figure 3A). The extraction and sequencing of RNA obtained from cells overexpressing circFGD4 confirmed that the overexpressed circFGD4 was circular rather than linear. A series of proliferation-related experiments indicated that the viability, colony formation, and DNA synthesis abilities of MGC803 and SGC7901 cells were significantly decreased in the circFGD4 overexpression group compared with the negative control (NC) group (Figure 3B–D). Transwell migration assays demonstrated that the migrated cells were significantly decreased when circFGD4 was overexpressed at MGC803 and SGC7901 cells (Figure 3E). Also, IF assays demonstrated that the gain of circFGD4 condensed and aggregated the morphology of MGC803 cells without dispersion, followed by changes in EMT markers, including an increased E-cadherin expression and decreased vimentin expression (Figure 3F). In contrast, EMT changes after enforced circFGD4 were opposite to the effect of knocking down circFGD4 in GC cells. The data confirmed that circFGD4 was a tumour suppressor of GC progression and metastasis in vitro.

To elucidate whether circFGD4 exerts anticancer effects through sponge-binding miRNAs, we first predicted a total of 55 potential conjugated miRNAs using online prediction tools (Circinteractome [15], Circbank [16], RegRNA [17], CSCD [18]; Figure 4A). The qRT-PCR results indicated that the expression levels of hsa-miR-32-5p, hsa-miR-150-5p, has-miR-532-3p, and hsa-miR-629-3p in 20 GC tissues compared with their paired noncancerous tissues (Figure 4B). Further validation revealed that miR-532-3p expression was higher in GC tissues than in noncancerous tissues (Figure 4C). FISH assays were used, and the results indicated that circFGD4 and miR-532-3p were colocalized in the cytoplasm of MGC803 and AGS cells (Figure 4D). Subsequently, the circFGD4-WT and circFGD4-MUT miR-532-3p binding sites were subcloned into the dual-luciferase reporters (Figure 4E). The results revealed that miR-532-3p mimics reduced the luciferase activity of circFGD4-WT, but had no effects on that of circFGD4-MUT in HEK-293T cells (Figure 4F). Additionally, an anti-AGO2 RIP assay revealed that endogenous circFGD4 was pulled down by anti-AGO2 antibodies and was mainly enriched in the AGO2 group than the IgG control group (Figure 4G), and both circFGD4 and miR-532-3p were also predominantly enriched in AGS cells which were transfected with miR-532-3p mimics (Figure 4H). Next, by using circFGD4 pull-down experiments, we purified circFGD4-associated RNAs and analysed candidate miRNAs in the complex. The specific enrichment of miR-532-3p that was abundantly detected by qRT-qPCR in the circFGD4 probe and was markedly higher than the miR-NC group in AGS cells (Figure 4I). Furthermore, the expression of miR-532-3p was reversely correlated with that of circFGD4 in GC tissues (r = −0.412, P=0.0407; Figure 4J). These results demonstrated that circFGD4 acted as an effective sponge of miR-532-3p in GC.

We further validated the functional roles of circFGD4 and miR-532-3p in GC cells by rescue assays such as CCK-8, colony formation, and Transwell migration assays. The levels of miR-532-3p were up- or down-regulated by qRT-PCR after the transfection of MGC803 or AGS cells with miR-532-3p mimics or inhibitors, respectively (Figure 5A). CCK-8 assays indicated that circFGD4 inhibited MGC803 cell proliferation and that miR-532-3p mimics attenuated this inhibition. Also, the silencing of circFGD4 promoted AGS cell growth, and the miR-532-3p inhibitors relieved this promotion (Figure 5B). Western blotting indicated that circFGD4 decreased the PCNA and Ki-67 expression at protein levels, and miR-532-3p mimics inversed this tendency (Figure 5C). Colony formation assays showed similar proliferation effects (Figure 5D). Transwell assays validated that circFGD4 reduced MGC803 cell migration and that the miR-532-3p mimics attenuated this suppression. Silencing of circFGD4 increased AGS cell migration, and miR-532-3p inhibitors weakened this acceleration (Figure 5E). Western blotting demonstrated that the circFGD4/miR-532-3p axis also regulated the EMT-related markers (such as E-cadherin, N-cadherin, and vimentin) in GC cells (Figure 5F). These results indicated that circFGD4 inhibited GC progression and metastasis by abolishing the oncogenic effect of miR-532-3p.

Based on the combined results of target prediction analyses, we found that circFGD4 and the 3′-UTR of APC shared the same miR-532-3p (miRNA Response Elements) MREs (Figure 5G). Western blotting verified that overexpression or silencing of circFGD4 increased or reduced the expression of APC, respectively. Also, miR-532-3p mimics and inhibitors regulated the expression of the APC mRNA and protein, respectively. Rescue assays further indicated that the up- or down-regulation of APC expression levels transfected with miR-532-3p inhibitors or mimics could be restored by the cotransfection of shRNA1 + miR-532-3p inhibitors or circFGD4 overexpression + miR-532-3p mimics (Figure 5H). Subsequently, full-length APC-WT and APC-MUT (red font as mutant locus) were subcloned into the dual-luciferase reporters (Figure 5I). As expected, the luciferase reporter gene assays demonstrated that cotransfection with miR-532-3p mimics and pGL3-APC-WT vector exhibited reduced luciferase activity, whereas miR-532-3p mimics failed to inhibit the luciferase activity of pGL3-APC-MUT vector in HEK-293T cells (Figure 5J). Additionally, the bioinformatics analysis indicated that GC patients with low APC expression had a poor overall survival compared with those with high APC expression (Figure 5K). IF assays also revealed that circFGD4 significantly controlled the expression of APC via the regulation of miR-532-3p in MGC803 and AGS cells (Figure 5L). All these data indicated that circFGD4 could affect the expression of APC by directly interacted with miR-532-3p in GC.

A multi-database analysis found that multiple mRNAs, including APC, were potential targets of miR-532-3p (Figure 6A). The underlying biological processes of 159 potential target mRNAs of miR-532-3p were predicted by the DAVID v6.8 database, and the results indicated that the downstream genes were preferentially enriched in cell migration and the Wnt/β-catenin pathway (Figure 6B,C). Meanwhile, APC protein, a key negative regulator, contributed to multiple tumours progression by modulating Wnt/β-catenin signalling pathway [19]. Therefore, TOP/Flash and FOP/Flash reporter gene were constructed and performed to detect whether the Wnt/β-catenin pathway was activated in GC cells. TOP/FOP transcriptional activity was enhanced or reduced after the transfection of the knockdown or overexpression of circFGD4 vectors in HEK-293T, MGC803, and AGS cells (Figure 6D,E). Further experiments showed that circFGD4 overexpression with the transfection of MGC803 and AGS cells reversed the increase in TOP/FOP transcriptional activity induced by miR-532-3p mimics (Figure 6F). The knocking down of circFGD4 also reversed the decline in the TOP/FOP transcriptional activity prompted by miR-532-3p inhibitors in MGC803 and AGS cells (Figure 6F). Western blotting indicated that cotransfection of circFGD4 and miR-532-3p in MGC803 and AGS cells could not significantly modulate the total protein of β-catenin; However, β-catenin protein in the nucleus of GC cells displayed distinct changes (Figure 6G). The nuclear/cytoplasmic ratios of β-catenin were significantly increased in shRNA1-transfected AGS cells compared to AGS cells transfected only with miR-532-3p inhibitors, but cotransfection of both reduced this difference (Figure 6G). Also, similar nuclear/cytoplasmic ratios of β-catenin were examined in MGC803 cells transfected with miR-532-3p mimics and/or circFGD4 overexpressing vector (Figure 6G). These results indicated that circFGD4 decreased the aggregation of β-catenin in the nucleus of GC cells by suppressing miR-532-3p and facilitating APC expression, thereby inactivating the classical Wnt/β-catenin signalling pathway (Figure 6H).

To explore the roles of circFGD4 on tumour growth, we established a mouse xenograft model. The nude mice were killed after 4 weeks; Xenograft tumours were excised and photographed (Figure 7A). In contrast with the negative control vector (NC) group, the xenograft tumours in the circFGD4 overexpression group indicated a decrease in volume and weight (Figure 7B,C). We conducted qRT-PCR assay on the collected tumours; the results showed that circFGD4 overexpression significantly reduced mRNA levels of miR-532-3p, Ki-67, PCNA, and Vimentin expression and increased APC and E-cadherin expression (Figure 7D). IHC assay was further performed; the results indicated that circFGD4 overexpression enhanced the immunochemical staining of APC and E-cadherin and reduced the immunostaining of Ki-67, PCNA, and Vimentin (Figure 7E). After that, to construct a lung metastasis model, the number and size of lung metastatic nodules were significantly lower in nude mouse injected by MGC803 cells with circFGD4 overexpressing group compared with that nude mouse injected with cells from blank or NC group (Figure 7F). Also, the mRNA levels of miR-532-3p, Ki-67, PCNA, and Vimentin in the circFGD4 overexpression group were lower than that in the NC group, while the APC, E-cadherin level was increased in circFGD4 overexpression group (Figure 7G). These data indicated that circFGD4 suppressed the tumorigenesis and metastasis of GC in vivo.

A large number of circRNAs play essential roles as regulators and are valuable diagnostic markers for tumorigenesis and tumour progression [20]. For example, down-regulation of hsa_circ_0009361 has been detected in the cancerous tissues from colorectal cancer patients (CRC), and its use as a biomarker has been suggested, as low expression of hsa_circ_0009361 predicted cancer progression, metastasis and induced EMT of CRC patients [21]. Circular RNA cSMARCA5 is highly down-regulated in the tissues resected from hepatocellular cancer (HCC), and cSMARCA5 is correlated with advanced tumour stage and poor prognosis of HCC patients [22]. Here, we performed a circRNA microarray screen and selected circFGD4 to study its biological function further. We validated that circFGD4 was prominently reduced in GC tissues and cell lines and that lower expression of circFGD4 was markedly correlated with poor differentiation, lymphatic metastasis, and poor prognosis of GC patients. Functional experiments confirmed that circFGD4 inhibited tumour growth, metastasis, and induced EMT in GC cells. These data indicated that circFGD4 could act as a tumour suppressor in the tumorigenesis and metastasis of GC and might be a novel potential biomarker and therapeutic strategy for GC.

A growing number of studies have shown that cytoplasmic ncRNAs, including circRNAs, lncRNAs, and miRNAs, can affect mRNA stability and translation process or can regulate signalling pathways through competing with miRNAs for shared MREs [23–25]. In our study, preliminary bioinformatics analysis indicated that circFGD4 contained the MRE of miR-532-3p. Thus, we presumed that circFGD4 might play an anti-cancer effect through sponging miR-532-3p in GC. Subsequently, RNA FISH, dual-luciferase reporter gene, RNA immunoprecipitation, and circFGD4 pull-down assays demonstrated that circFGD4 interacted directly with miR-532-3p in GC. The rescue experiments further verified that circFGD4 inhibited tumour progression and metastasis by directly sponging miR-532 in GC. Next, we found that the APC gene, vital cell proliferation, and migration regulator, had a potential binding site for miR-532-3p according to bioinformatics analysis. We observed that circFGD4 not only regulated the expression level of APC in GC cells but that miR-532-3p also affected the expression of APC at the mRNA and protein levels, suggesting that the APC gene could probably be the target molecule of the circFGD4/miR-532-3p axis. Furthermore, dual-luciferase reporter assays determined that miR-532-3p could directly bind to APC. Western blotting and IF analysis indicated that circFGD4 could regulate the expression of APC via miR-532-3p, suggesting that circFGD4 inhibits GC progression through sponging miR-532-3p and thus regulating APC gene expression in GC cells.

Many works of literature have demonstrated that the Wnt/β-catenin signalling pathway is closely related to the viability, metastasis, and EMT of many malignant tumours [26,27]. The inhibition of the Wnt/β-catenin pathway by APC protein has been demonstrated in many studies [28,29]. Based on bioinformatics analysis, we found that cell migration and the Wnt signalling pathway were biological processes and pathways of miR-532-3p targeting genes. We further found that circFGD4 overexpression reduced the TOP/FOP activity, while knockdown of circFGD4 had the reverse effect; these results suggested that the Wnt/β-catenin signalling pathway was inactivated. Based on the further studies, increasing the miR-532-3p expression could significantly enhance the TOP/FOP transcriptional activity, suggesting that the Wnt/β-catenin signalling pathway was activated and cotransfection with circFGD4 shRNA1/circFGD4 overexpressed vector and miR-532-3p inhibitors/ mimics can enhance or reduce the TOP/FOP activity. Therefore, our data indicated that a novel circFGD4/miR-532-3p/APC regulatory network suppressed the progression and metastasis of GC by inactivating the β-catenin signalling.

In summary, we demonstrated that circFGD4, which directly targets miR-532-3p/APC, suppressed GC cell viability, metastasis, and induced EMT through inactivating the β-catenin signalling. Our study suggests that circFGD4 serves a novel biomarker and therapeutic strategy for anti-GC therapy.

• The roles of circRNAs in multiple cancers are not fully understood, and the underlying mechanisms by which circFGD4 contributed to gastric cancer (GC) remain mostly uncovered.

• We identified that circFGD4 was markedly down-regulated in GC, and lower expression of circFGD4 was correlated with the poor tumour differentiation, lymph node metastasis, and poor prognosis of GC patients. The low expression level of circFGD4 promoted the viability, migration, and induced EMT of GC cells. The mechanistic analysis demonstrated that circFGD4 suppressed GC progression through sponging miR-532-3p and enhancing APC expression to inactivate the β-catenin signalling.

• We identified a novel functional pathway controlled by the circFGD4/miR-532-3p/APC network. This regulatory network may provide a novel potential biomarker and valuable therapeutic strategy for GC.

All the data and materials supporting the conclusion of the present study have been included within the article.

The authors declare that there are no competing interests associated with the manuscript.

This work was supported by the National Natural Science Foundation of China [grant number 81974385] and Chongqing Municipal Key Discipline Funding from the funding agencies of Chongqing Municipal Education Commission [grant number 201128GRJFJY].

Xinglong Dai and Jianjun Liu substantially contributed to the conception and design of the work; Xiong Guo, Xiaoya Deng, and Anqi Cheng contributed to acquisition, analysis and interpretation of the data; Xinglong Dai drafted the manuscript; Xinglong Dai, Jianjun Liu, Xiong Guo, Anqi Cheng, Xiaoya Deng, Liqun Guo, and Ziwei Wang revising the work critically; all authors gave final approval of the work.

The Ethics Committee of the First Affiliated Hospital of Chongqing Medical University approved the study, and all participants signed informed consent statements. All animal experiments were approved by the Animal Ethics Committee of Chongqing Medical University.

APC

adenomatous polyposis coli

CCK-8

Cell Counting Kit-8

circFGD4

circular RNA FGD4

circRNA

circular RNA

DAB

diaminoaniline

DAPI

4′,6-Diamidino-2-phenylindole

EdU

5-Ethynyl-2′-deoxyuridine

EMT

epithelial–mesenchymal transition

FISH

fluorescence in situ hybridization

GAPDH

glyceraldehyde 3-phosphate dehydrogenase

GC

gastric cancer

HRP

horseradish peroxidase

IF

immunofluorescence

IHC

immunohistochemistry

MREs

miRNA response elements

MUT

mutant-type

PVDF

polyvinylidene fluoride

qRT-PCR

quantitative real-time PCR

RIP

RNA immunoprecipitation

UTR

untranslated region

WT

wild-type